Add host commandline argument for UI and sklearn server

[mdagost]

The UI and the model server run on 127.0.0.1 by default, but this doesn't work from within a docker container. This PR introduces a -host commandline flag so that you can pass in 0.0.0.0 to listen on all interfaces.

To verify that this works, you can do the following:

git clone https://github.com/databricks/mlflow.git
cd mlflow
docker build -t mlfow .
docker run -it -p 5000:5000 mlflow

Then from within the container:

mlflow ui

And go to http://127.0.0.1:5000 and notice that you don't see the UI. Then from within the container instead do:

mlflow ui --host=0.0.0.0

and visit http://0.0.0.0:5000

To see the model serving stuff, go into the container and do

python example/quickstart/test_sklearn.py

and note the model id. Then do

mlflow sklearn serve -r <MODEL_ID> model

and from a separate terminal try (but fail) to get a prediction:

curl -d '[{"x": 1}, {"x": -1}]' -H 'Content-Type: application/json'  -X POST localhost:5000/invocations

Instead try

 mlflow sklearn serve -r 03947ea706cd474aa4d2395db7d9d6ff --host=0.0.0.0 model

curl -d '[{"x": 1}, {"x": -1}]' -H 'Content-Type: application/json'  -X POST 0.0.0.0:5000/invocations

[aarondav]

Thanks for the contribution! This looks great. Would you mind adding a PR description to give some context about why this change is made, including the steps you took to verify that this works as expected?

[aarondav]

Looks like this line (and the associated one) has a linter error due to line length. In general you can run bash lint.sh locally to run the linter.

Maybe we could change the help a bit while we're at it (mostly to clarify the implication of the default and of setting 0.0.0.0):

@click.option("--host", default="127.0.0.1",
              help="The IP address which the UI server listens."
                   " Defaults to 127.0.0.1, allowing only local connections."
                   " Use 0.0.0.0 to bind to all addresses.")

[aarondav]

Maybe here instead of "UI server" we can say "prediction server".

[mdagost]

@aarondav I added a PR description and addressed your help text (and linting) comments.

[codecov-io]

Codecov Report

Merging #27 into master will increase coverage by 0.1%.
The diff coverage is 40%.

@@            Coverage Diff            @@
##           master      #27     +/-   ##
=========================================
+ Coverage    69.6%   69.71%   +0.1%     
=========================================
  Files          44       44             
  Lines        2527     2536      +9     
=========================================
+ Hits         1759     1768      +9     
  Misses        768      768

Impacted Files	Coverage Δ
mlflow/cli.py	0% <0%> (ø)	⬆️
mlflow/sklearn.py	78.68% <66.66%> (+0.35%)	⬆️
mlflow/store/file_store.py	90.95% <0%> (+0.08%)	⬆️
mlflow/projects.py	54.22% <0%> (+0.2%)	⬆️
mlflow/tracking/__init__.py	82.22% <0%> (+0.4%)	⬆️
mlflow/version.py	100% <0%> (+100%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 13bbd76...138b55e. Read the comment docs.

[aarondav]

LGTM - merging!